Hate Speech
and Offensive Content Identification
in Memes

This shared task focuses on the automatic identification of hate speech, offensive content and other harmful attributes in memes. The task covers Tamil and Telugu languages and aims to advance research on multilingual, multimodal content understanding.

Social media platforms have become a battleground for misinformation and hateful content. Memes, due to their inherently multimodal nature, present a unique challenge: text alone may appear benign, images may seem neutral, yet together they can convey deeply offensive or hateful messages. Research has shown hateful memes can be surprisingly persuasive, often going viral and resonating with a wide audience.

This task is part of FIRE 2026 (Forum for Information Retrieval Evaluation) and invites researchers to develop automated systems capable of detecting and classifying hate speech in memes, pushing the frontier of multilingual and multimodal NLP. Each meme entry also includes a context paragraph to aid analysis.

Participants address five subtasks: abuse detection, target community identification, vulgarity detection, sarcasm detection, and sentiment classification. Accepted system description papers will be published in FIRE 2026 proceedings.